Teaching Block Series
Eight teaching blocks documenting how session-bound AI became a filesystem-native ambient intelligence system — with a global skill library, forked writing pipelines, model-effort routing, and isolated catalog discovery.
Video
The complete experiment narrated end-to-end — session-bound AI to filesystem-native ambient intelligence, forked writing pipelines, model-effort routing, and catalog discovery.
Watch →Articles
How a short definition of ambient intelligence evolved into a filesystem-native architecture with manifests, inherited capabilities, and startup hooks.
Read →Why presence isn't execution — and how a global session-start hook solved what a document couldn't.
Read →How a forked writing pipeline became a reusable capability architect — and how model-effort routing got extracted into its own process.
Read →Auditing a model-selection guide against the actual pipeline that would run it — and what the code revealed.
Read →Separating model/effort routing from capability construction — and why that separation matters at scale.
Read →How a design contradiction led to isolated discovery agents — and why a disposable context is better than a sharded catalog.
Read →Running the writing-team-orchestrator end-to-end — interview, research, forks, and a finished article.
Read →Converting an exploratory model-comparison session into a shareable AIMM guide — and updating it when the operating surface changed.
Read →